Panasonic DC-GH6 Sensor Measurements at PhotonsToPhotos

Ranlee wrote:

Tim200 wrote:

Simon97 wrote:

I look at the studio scene RAW conversions. Ask yourself if the difference between models is really that drastic. What if you scale the 25mp images down to 20 and compare to the OM-1 or other MFT cameras? Would it make a difference in your photography?

I'd get the GH6 if I had a collection of MFT lenses and wanted to get more into video. Looking at the sample gallery it is a pretty good photo shooter but would lean to the OM-1 for straight photo work.

Yes, the gh6 and om1 look about equal in the studio scene. In fact I think gh6(scaled) actually looks a bit better at higher ISO e.g. 6400. In my use, stills look fine as long as I'm careful with how the shot is exposed. However, what I'm also seeing is the issue outlined in the dpreview article about gh6 low ISO DR. If I have to raise shadows, it quickly becomes noisy. Example I've run into is a subject in front of a bright window. It has taken some practice to get a shot like this which can be processed decently.

This is a reasonable statement and I certainly am not qualified to dispute it. However, the question for me is compared to what? I have taken test shots like you describe (bright window dark shadows) with both G9 and GH6 with exposure as close as I could get it -1.3 GH6 -1G9 (I think the GH6 meters brighter than the G9, even a third stop wasn't quite enough). I imported the Raw's to LRC, zeroed sharpening and NR, brightened by +2 and exported. Regarding noise, in every case from ISO100 to 25600 I would take the GH6 files over the G9's. As far DR, I could not see a difference in my test files, and the G9 is regarded as best in MFT for DR. Now, was this scientific - hardly. But it was done as equally as I could within my expertise, and within the parameters of my normal shooting. For example, I didn't raise anything by 5ev and probably never will.

Now, I expect someone will tell me what I did wrong in testing. That's cool, but doesn't mean I will understand them or suddenly decide my GH6 is trash. So far I happen to like the camera.

That's interesting, and good to hear. It's good to see some positive reports from practical use.  I've shot full frame the last couple of years, so that's probably slanting my perspective a bit. Despite the scenario I mentioned, I love the gh6 and don't regret getting it.

Dale108 wrote:

Hi Randy:

Test charts are fascinating but real world shooting is what matters in the end.

Have you tested GH6 CAF vs G9? Curious if any improvements.

Dale

Hi Dale.  As far as moving subjects and CAF I simply haven't tested.  Semi-stationary wildlife and landscapes are mostly where I dwell but I can give you my impressions of the focusing.

It's been a cold, rainy, snowy spring so far which has limited me to my yard for the most part.  I had lot of trouble in the beginning shooting a few Robins and a rabbit that doesn't mind posing.  It was frustrating as I couldn't to get many shots in focus.  I tried all the different modes with little in the way of results.  It dawned on me that I was shooting both through an open window and downward, and from too far a distance.  The Robins always had grass etc.. close behind and the Rabbit had drooping vegetation in front and strong contrasty objects behind.  I was continually getting focus behind the subjects.  I got quite a bit of improvement using SAF and using the Near focus option.  I got even better results simply by going out of the house and shooting from a different angle.  So it was me and the conditions that were the problem.  I'm getting my usual high percentage of in focus shots now in SAF with the small focus box.

Regarding CAF.  I rarely used it on my G9.  The constant warbling in the view finder was distracting and I didn't find it that accurate.  I have been using CAF quite a bit on the GH6 and find the small warble doesn't bother me like it does on the G9.  Probably because of the improved speed I suppose.  I have using it the same as do with SAF, that is single small focus box and am getting consistently good results.  I have not had much luck with the animal detection to date.   It finds the animal easily enough and consistently, But high contrast backgrounds and the box being always slightly larger than the animal cause more rear focusing.  This is mitigated somewhat by the Near option again but not very satisfactorily.  Hopefully when I get back to my woods I'll be able to vary the backgrounds and do a better test.  I do think I will be using CAF a lot more with the GH6.

I do really miss the AF Focus Scope.  I use it all the time on my G9.  I hope a firmware update will give that back.

Something else I found (nothing to do with AF).  The option to save settings to the memory card now includes an option to have formatting leave the file alone.  I like this.  On the G9 I usually do a quick erase of files by using the format command which always also removed the settings file.  Ok, I got a little windy.  Hope this was helpful.

Thanks Randy

Dale

Ranlee wrote:

Dale108 wrote:

Hi Randy:

Test charts are fascinating but real world shooting is what matters in the end.

Have you tested GH6 CAF vs G9? Curious if any improvements.

Dale

Hi Dale. As far as moving subjects and CAF I simply haven't tested. Semi-stationary wildlife and landscapes are mostly where I dwell but I can give you my impressions of the focusing.

It's been a cold, rainy, snowy spring so far which has limited me to my yard for the most part. I had lot of trouble in the beginning shooting a few Robins and a rabbit that doesn't mind posing. It was frustrating as I couldn't to get many shots in focus. I tried all the different modes with little in the way of results. It dawned on me that I was shooting both through an open window and downward, and from too far a distance. The Robins always had grass etc.. close behind and the Rabbit had drooping vegetation in front and strong contrasty objects behind. I was continually getting focus behind the subjects. I got quite a bit of improvement using SAF and using the Near focus option. I got even better results simply by going out of the house and shooting from a different angle. So it was me and the conditions that were the problem. I'm getting my usual high percentage of in focus shots now in SAF with the small focus box.

Regarding CAF. I rarely used it on my G9. The constant warbling in the view finder was distracting and I didn't find it that accurate. I have been using CAF quite a bit on the GH6 and find the small warble doesn't bother me like it does on the G9. Probably because of the improved speed I suppose. I have using it the same as do with SAF, that is single small focus box and am getting consistently good results. I have not had much luck with the animal detection to date. It finds the animal easily enough and consistently, But high contrast backgrounds and the box being always slightly larger than the animal cause more rear focusing. This is mitigated somewhat by the Near option again but not very satisfactorily. Hopefully when I get back to my woods I'll be able to vary the backgrounds and do a better test. I do think I will be using CAF a lot more with the GH6.

I do really miss the AF Focus Scope. I use it all the time on my G9. I hope a firmware update will give that back.

Something else I found (nothing to do with AF). The option to save settings to the memory card now includes an option to have formatting leave the file alone. I like this. On the G9 I usually do a quick erase of files by using the format command which always also removed the settings file. Ok, I got a little windy. Hope this was helpful.

knickerhawk wrote:

Interceptor121 wrote:

drj3 wrote:

It looks the same to me, though we have no measurements.

On the M1x, tripod HR takes only 8 shots and has a very similar PDR to HHHR (16-shots). We do not know how many of those 16 or 12 shots are used by the HHHR algorithm. GH6's HHHR mode picks the best 8 out of 16 shots.

The OM-1 and probably the E-M1s use 8 of the images. I assume the OM-1 can get the same IQ with fewer images because of the faster readout speed of the sensor and faster frame rates which would reduce the chances of other factors affecting the images.

You can count the number used by looking at an image with something moving (like counting the number of London has fallen in the attached which was scrolling on a TV screen (7 plus the one blank).

I could post an image of a waterfall where you see the many hundreds of droplets 8 times, but I would have to process the image and load it over my slow internet connection.

I think what is likely to happen here is simply that is using the first set of 8 images and then 4 of the 8 for the corners where it is possibly getting most of the issues for a variety of reasons including distortion.

If it has more speed it would not make sense to drop any frames unless they have worked out that some of them do not do much statistically

But as all of this is not clearly explained we will not know

At least with respect to the EM1iii, HHHR definitely drops varying numbers of frames. It also frequently fails completely and reports an error and doesn't record an HHHR image at all. Here's a test I did with a swinging flashlight. About 2/3 of the HHHRs successfully recorded used all 16 shots, the other 1/3 of the HHHRs used fewer than 16 shots.

Top=All 16 shots recorded; Upper Middle=15 shots used with 5th (counting from left) missing; Lower Middle=14 shots used with 5th and 9th (counting from left) missing; Bottom=13 shots used with 5th, 9th and 15th (counting from the left) missing.

Thanks for that it gives a great visual of what is going on

knickerhawk wrote:

Interceptor121 wrote:

drj3 wrote:

It looks the same to me, though we have no measurements.

On the M1x, tripod HR takes only 8 shots and has a very similar PDR to HHHR (16-shots). We do not know how many of those 16 or 12 shots are used by the HHHR algorithm. GH6's HHHR mode picks the best 8 out of 16 shots.

The OM-1 and probably the E-M1s use 8 of the images. I assume the OM-1 can get the same IQ with fewer images because of the faster readout speed of the sensor and faster frame rates which would reduce the chances of other factors affecting the images.

You can count the number used by looking at an image with something moving (like counting the number of London has fallen in the attached which was scrolling on a TV screen (7 plus the one blank).

I could post an image of a waterfall where you see the many hundreds of droplets 8 times, but I would have to process the image and load it over my slow internet connection.

I think what is likely to happen here is simply that is using the first set of 8 images and then 4 of the 8 for the corners where it is possibly getting most of the issues for a variety of reasons including distortion.

If it has more speed it would not make sense to drop any frames unless they have worked out that some of them do not do much statistically

But as all of this is not clearly explained we will not know

At least with respect to the EM1iii, HHHR definitely drops varying numbers of frames. It also frequently fails completely and reports an error and doesn't record an HHHR image at all. Here's a test I did with a swinging flashlight. About 2/3 of the HHHRs successfully recorded used all 16 shots, the other 1/3 of the HHHRs used fewer than 16 shots.

Top=All 16 shots recorded; Upper Middle=15 shots used with 5th (counting from left) missing; Lower Middle=14 shots used with 5th and 9th (counting from left) missing; Bottom=13 shots used with 5th, 9th and 15th (counting from the left) missing.

The HHHR of the GH6 can also fail when not exposed correctly

Interceptor121 wrote:

JakeJY wrote:

Interceptor121 wrote:

JakeJY wrote:

dbateman wrote:

Funny Valentine wrote:

hmmm, what a disaster for Panasonic... GH6 scores worse than the OM-1 and and last gen EM1-2 and 3...

You don't seem to know how to read that graph. This is quite good for a very fast front side illuminated higher resolution sensor optimized for video. If anything the OM-1 isn't looking quite as good as it should. The Em1mk2 needs to be corrected to the right 1/3 a stop due to gain choices. The stacked nature of the Om-1 semsor is hurting significantly any gains from being BSI, and its still very slow read out sensor compared to even the GH6.

I can see why now Panasonic choose this sensor over the Sony stacked BSI one. Its better.

FYI the sensor readout of the OM1 is 120fps, the fastest in M43. The GH6 is only 75fps.

Readout in fps depends on bit depth and pixel count

120x20.2x12=29088

75x25x16=30000

Under ISO 800 it doesn't look to have anything that requires 16 bit. From the results, it looks like that would even fit in 10 bit (the results from latitude test are consistent with the G95 in 10-bit mode, perhaps even a bit worse).

If you drop the bit depth to 12 bits you get 100 fps at 25 megapixels if you bin downwards you get to the same point

The camera have different objectives and design principles and can't really be compared

When it's doing the dual output gain the stream into the ISP may be even more than 16 bits total, but that is hardly relevant to the max fps the sensor can do. I doubt your claim it can do 100 fps at 25MP in 12 bit, as Panasonic does not offer that mode at all.

All video modes are read at 12 bits and scaled to 10 in the process.

The camera is limited at the compression stage not readout for video

Note I'm talking purely about the sensor here and not only about video, also including the RAW stills. Currently Panasonic only offers 75 fps mode as the max for both full res stills and video regardless of ending bitrate. They don't offer a 100 fps mode full res mode by dropping to 12 bit or even 10 bit. Compression bottlenecks can be addressed in future cameras by ISP upgrades, but sensor bottlenecks will always be there.

OM1 instead offers 120 fps outright at 12 bit, so we can be sure the sensor at minimum supports that.

JakeJY wrote:

Interceptor121 wrote:

JakeJY wrote:

Interceptor121 wrote:

JakeJY wrote:

dbateman wrote:

Funny Valentine wrote:

hmmm, what a disaster for Panasonic... GH6 scores worse than the OM-1 and and last gen EM1-2 and 3...

You don't seem to know how to read that graph. This is quite good for a very fast front side illuminated higher resolution sensor optimized for video. If anything the OM-1 isn't looking quite as good as it should. The Em1mk2 needs to be corrected to the right 1/3 a stop due to gain choices. The stacked nature of the Om-1 semsor is hurting significantly any gains from being BSI, and its still very slow read out sensor compared to even the GH6.

I can see why now Panasonic choose this sensor over the Sony stacked BSI one. Its better.

FYI the sensor readout of the OM1 is 120fps, the fastest in M43. The GH6 is only 75fps.

Readout in fps depends on bit depth and pixel count

120x20.2x12=29088

75x25x16=30000

Under ISO 800 it doesn't look to have anything that requires 16 bit. From the results, it looks like that would even fit in 10 bit (the results from latitude test are consistent with the G95 in 10-bit mode, perhaps even a bit worse).

If you drop the bit depth to 12 bits you get 100 fps at 25 megapixels if you bin downwards you get to the same point

The camera have different objectives and design principles and can't really be compared

When it's doing the dual output gain the stream into the ISP may be even more than 16 bits total, but that is hardly relevant to the max fps the sensor can do. I doubt your claim it can do 100 fps at 25MP in 12 bit, as Panasonic does not offer that mode at all.

All video modes are read at 12 bits and scaled to 10 in the process.

The camera is limited at the compression stage not readout for video

Note I'm talking purely about the sensor here and not only about video, also including the RAW stills. Currently Panasonic only offers 75 fps mode as the max for both full res stills and video regardless of ending bitrate. They don't offer a 100 fps mode full res mode by dropping to 12 bit or even 10 bit. Compression bottlenecks can be addressed in future cameras by ISP upgrades, but sensor bottlenecks will always be there.

OM1 instead offers 120 fps outright at 12 bit, so we can be sure the sensor at minimum supports that.

As I said the limit is on the output

the OM-1 may run at 120 fps but other than writing to a buffer for less than 1 second  no function runs at that speed autofocus works at 60fps so in effect there is no tangible benefit for a use case

the 75fps of the GH6 is also not that useful to be frank

in order to run the camera at 120 fps you need cooling the 4k120 fps mode is the only one that gets the camera hot

instagram http://instagram.com/interceptor121
My flickr sets http://www.flickr.com/photos/interceptor121/
Youtube channel http://www.youtube.com/interceptor121
Underwater Photo and Video Blog http://interceptor121.com
Deer Photography workshops https://interceptor121.com/2021/09/26/2021-22-deer-photography-workshops-in-woburn/

knickerhawk wrote:

Raist3d wrote:

Funny Valentine wrote:

Raist3d wrote:

Funny Valentine wrote:

Pete Berry wrote:

And that reminds me: my almost 5 year old G9 sensor still rules the m4/3 DR roost along with it's use in the more recent GH5-II if you're into the top-dog sensor thing... I do like the larger GH6 stills files, along with the excellent HHHR implementation.

Somehow it seems the GH6 HHHRS doesn't suppress noise like Olympus HHHRS. This needs to be confirmed by someone who owns both.

Also we need to know if the OM-1 HHHRS is as good as E-M1 III HHHRS because they went down to 12 stacked photos from 16. Why would they do this ?

Because due to faster cpu/sensor readout, they can be more confident the number of images they need to create the HHHR are less than the other case, where they may have had to throw away more of the images.

according to Robin Wong and Olympus themselves, in the E-M1 III and X, the HHHRS RAW files were made of 16 stacked photos, for a total of 320 megapixels that were oversampled to 50 mp.

They do take 16 shots but are they using them? And I wouldn't believe everything Robin says. He's not perfect - like for example saying the EM5.3 EVF can do 120fps refresh rate- that's not true.

See my post here. At least with respect to the EM1iii, often all 16 shots are used in the HHHR image generated by the camera, but sometimes its fewer than that and sometimes the camera gives up and errors out instead of generating an HHHR image.

So if the math adds up, the OM-1 must be oversampled from 240 MP only, with 1 stop less DR than the E-M1 X and III.

Again, for hand held high res, you may not use all the shots. They need to register. WE need to confirm they are using all the shots. Also it's possible the new sensor with better color-in noise characteristics gives them the ability to use less shots.

So in ideal conditions, the old HHHRS should outperform the new HHHRS.

I wouldn't 'go by that assumption. A test should uncover this though.

Yes, it should be tested. Since the EM's can potentially incorporate more frames (and hence a larger aggregate exposure), they might have a modest noise advantage, but perhaps OM has tweaked the alignment algorithms or done other things to counter the need for the additional frames. Until tested, this is speculative.

So far, the OM1 has only given up on me once with HHHR, which is a marked improvement over the EM1iii.

In very difficult situations, the OM-1almost always succeeds in creating a stacked image, but the final image can be crooked when you really push it.

Pete Berry wrote:

bclaff wrote:

Funny Valentine wrote:

hmmm, what a disaster for Panasonic... GH6 scores worse than the OM-1 and and last gen EM1-2 and 3...

This camera is not designed with still photography in mind.

So to play Devil's advocate, Bill, 7 of the past 8 Panny m4/3s you've tested have shown higher PDR's than the OM-1 - as have all three of the Olys sharing the E-M1-III sensor, The G9 eclipses the OM-1 by a whopping 0.76 EV.

So surely the OM-1 was designed with only mediocre stills in mind, along with its sub-class video performance!

Having both the G9 and GH6 I've yet to experience any significant low-iso photographic difference in DR, and appreciate the 5MP frame increase.

the GH6 is such a mystery, there's little to no review about the photographic capabilities of that new 25mp sensor, only a bunch of spec sheet readers on youtube...

I would love to see a real life comparison between OM-1 and GH6 RAW files, using OM workspace and Silkypix to exploit them fully

Pete Berry wrote:

So to play Devil's advocate, Bill, 7 of the past 8 Panny m4/3s you've tested have shown higher PDR's than the OM-1 - as have all three of the Olys sharing the E-M1-III sensor, The G9 eclipses the OM-1 by a whopping 0.76 EV.

Not so much.  The 20MP Sony sensors had the same DR with little actual variance.  The "peak" you're looking comes at the expense of color accuracy as explained previously .

So surely the OM-1 was designed with only mediocre stills in mind, along with its sub-class video performance!

Obviously haven't used one.  It is the most complete M43 camera to date.

Having both the G9 and GH6 I've yet to experience any significant low-iso photographic difference in DR, and appreciate the 5MP frame increase.

Just don't lift the shadows and you'll be fine.

Pete Berry wrote:

That really is an excellent outcome Pete!

Pete Berry wrote:

AURA PA wrote:

Pete Berry wrote:

So to play Devil's advocate, Bill, 7 of the past 8 Panny m4/3s you've tested have shown higher PDR's than the OM-1 - as have all three of the Olys sharing the E-M1-III sensor, The G9 eclipses the OM-1 by a whopping 0.76 EV.

Not so much. The 20MP Sony sensors had the same DR with little actual variance. The "peak" you're looking comes at the expense of color accuracy as explained previously .

So surely the OM-1 was designed with only mediocre stills in mind, along with its sub-class video performance!

Obviously haven't used one. It is the most complete M43 camera to date.

Having both the G9 and GH6 I've yet to experience any significant low-iso photographic difference in DR, and appreciate the 5MP frame increase.

Just don't lift the shadows and you'll be fine.

Whatever...

Below, tight crop for 12x16" print from above. Shadows boosted 95% in ACR

Its a pretty bird but at 2MP with NR I’m not sure this says anything about shadow noise.

bclaff wrote:

Pete Berry wrote:

AURA PA wrote:

Pete Berry wrote:

So to play Devil's advocate, Bill, 7 of the past 8 Panny m4/3s you've tested have shown higher PDR's than the OM-1 - as have all three of the Olys sharing the E-M1-III sensor, The G9 eclipses the OM-1 by a whopping 0.76 EV.

Not so much. The 20MP Sony sensors had the same DR with little actual variance. The "peak" you're looking comes at the expense of color accuracy as explained previously .

So surely the OM-1 was designed with only mediocre stills in mind, along with its sub-class video performance!

Obviously haven't used one. It is the most complete M43 camera to date.

Having both the G9 and GH6 I've yet to experience any significant low-iso photographic difference in DR, and appreciate the 5MP frame increase.

Just don't lift the shadows and you'll be fine.

Whatever...

Below, tight crop for 12x16" print from above. Shadows boosted 95% in ACR

That's not a high dynamic range scene, dark cormorant with shadow on neck.

But push the neck hard and you get this.

The GH6 does not have good deep blacks this is what you see creeping here

A camera with good blacks will resist manipulations of shadow better than another with better DR but worse blacks

In the case of the GH6 I can say the GH5M2 can take a lot more beating and even more of the G9 as it has better blacks